I now have the David Reinagels 16bit Random Number algorithm -
thanks.

I am looking to generate a random number in the range 0-1, then 0-2,
then 0-3,...... upto 0-49.

 I figured the easiest way might be for range 0-n

if n<=15 sum first n+1 bits in rand (16 bit number)
if 16<=n<=27 take first 4 bits<0-3> of rand giving a number 0-15 and
then add sum of remaining n-15 bits starting from bit <4>
if 28<=n<=30 take bits <0-4> giving a number 0-31 and subtract
sum of 31-n bits starting from bit <5>
if 31<=n<=42 take bits <0-4> and add sum of n-31 bits from <5>

For 43<=n<=50 I will probably have to generate 2 random nos and add
(eg for 0-50 generate one in range 0-15 and one  0-35  and add).
I'd be OK from 53-73 as I can use <0-5> and then
add/substract sum of remaining unused bits) as above.

This seems intrinsically faster than doing multiply/divides to
generate a number in  range. I don't believe it unbalances
randomness. Anyone got any better ideas (particularly for 43-50)?
Is this a well know technique?

BTW My earlier problems with MPLAB were indeed dumb user!

Brian.
Brian Jones
Java Technology Centre
IBM Hursley